Technologies:
pre-spaCy NLP
CRFs, Gensim, NLTK, sklearn, fuzzywuzzy, bs4, UIMA, IBM Watson, DKPro, Google Ngram Viewer, knowledge graphs, RDF triples, semantic search for retail, Google Adwords
pre-BERT NLP
spaCy, Conversational AI, Rasa NLU, Tensorflow, Trax, Keras, Pytorch, distantly supervised inference of demographic and psychographic attributes with web traffic data for audience intelligence, Google Insights Finder
pre-ChatGPT NLP
transfer learning, few-shot learning, common crawl, Leipzig corpora, intent detection and slot filling, emotion classification and sentiment analysis for brand safety, Google Display & Video 360 (Integral Ad Science, DoubleVerify, Adloox)
postmodern NLP
cute prompts for everyone! little retrieval pipelines for big retrieval pipeline, expressive formalisms, proof nets, category theory, Google extracts publisher data from digital news sites with Bard and Gemini then discourages people from visiting the sites by making the data available in Google Knowledge Graph (Google Search Generative Experience).
NLP no longer software, part of the foundation of software itself.
We continue training NLP systems with techniques from CV because they are pretty good at populating knowledge graphs with RDF triples most of the time.
ChatGPT integrations snowball. Check out this sick move: [prompts Prophet to create a UI component for configuring an agent and placing it in a Crew]
Machine learning engineers begin chunking the vectors!
Résumé/CV: https://seanbethard.net/bethard_cv.pdf
https://seanbethard.net/bethard_cv.docx
E-mail: sean@seanbethard.net
